Updated daily175 verified stories
New frontier, open-weight, multimodal, reasoning, and specialist AI models—plus the evidence behind their capability claims.
Every story links to the strongest available evidence: original sources when available, otherwise clearly attributed reporting.
What happened, why it matters, and what to watch — no jargon tax.
When the signal is thin, we publish nothing rather than padding the feed.
A growing stream of verified perspectives for people who need to understand AI without chasing hype.
A new benchmark of 11,586 bilingual, diagram-based physics problems reports that the strongest of 18 tested multimodal language models achieved only 33.7% answer accuracy.
A new arXiv benchmark evaluates LLM agents on geospatial planning using maps, tools and local social-media posts. Its authors report a sharp decline on complex tasks, with a 40.2% pass rate.
Google DeepMind says it is testing a Gemini Flash Lite model in a cryptographically protected environment designed to keep both evaluation prompts and model weights private.
South Korea released detailed second-phase scores for its Sovereign AI Foundation Model Project, with SK Telecom ranking first overall and Motif Technologies leading one global benchmark.
Google says Gemini Omni 1.1 Flash gives developers more control over AI-generated video through scene extension, first-and-last-frame interpolation, faster low-resolution drafts, video references and 4K upscaling.
Amazon Bedrock now routes OpenAI GPT-5.6 Terra and Luna requests between AWS Regions in Mumbai and Hyderabad, allowing customers to scale inference while keeping prompts and outputs within India, according to AWS.
Anthropic says scientists worldwide will receive free or discounted access to Claude for one year, alongside expanded research credits and continued limits on some life-sciences uses.
Google says Gemini Notebook can now answer questions about eligible ebooks purchased through Google Play Books, with citations, summaries, quizzes and other generated materials based on their contents.
Barret Zoph, a former Thinking Machines co-founder and OpenAI executive, has taken a vice president of research role at Google, TechCrunch reports. Google said he will bring reinforcement-learning and post-training expertise to its Gemini work, but the report does not establish his precise remit or start date.
The South China Morning Post reports that Zhipu AI formally released its open-weight GLM-5.3-Flash model after a stealth trial on 100,000 domestically produced chips, with unusually high usage on OpenRouter and OpenCode.
A security researcher reports that a crafted website and archive caused Claude Code Opus 5 in Auto Mode to execute attacker-controlled code in lab tests. The reported success rates reached 60% to 80% in small five-run samples, while Anthropic’s separate fixed evaluation reported zero successful attacks.
Fortune reports that Hong Kong startup Votee AI is retraining open-weight models on Cantonese data for banks, universities and government departments, aiming to address gaps in mainstream AI systems’ local and cultural knowledge.
One useful briefing each week
Get the week’s verified AI news, original data, useful tools, learning picks, and fresh AI jobs.
Hiring an AI professional or launching a useful AI product? Put it in front of people who came here to learn and act.
Post an AI job Submit an AI tool